Acceleration methods for multi-physics compressible flow

Oren Peles, Eli Turkel

Research output: Contribution to journalArticlepeer-review

Abstract

In this work we investigate the Runge–Kutta (RK)/Implicit smoother scheme as a convergence accelerator for complex multi-physics flow problems including turbulent, reactive and also two-phase flows. The flows considered are subsonic, transonic and supersonic flows in complex geometries, and also can be either steady or unsteady flows. All of these problems are considered to be a very stiff. We then introduce an acceleration method for the compressible Navier–Stokes equations. We start with the multigrid method for pure subsonic flow, including reactive flows. We then add the Rossow–Swanson–Turkel RK/Implicit smoother that enables performing all these complex flow simulations with a reasonable CFL number. We next discuss the RK/Implicit smoother for time dependent problem and also for low Mach numbers. The preconditioner includes an intrinsic low Mach number treatment inside the smoother operator. We also develop a modified Roe scheme with a corresponding flux Jacobian matrix. We then give the extension of the method for real gas and reactive flow. Reactive flows are governed by a system of inhomogeneous Navier–Stokes equations with very stiff source terms. The extension of the RK/Implicit smoother requires an approximation of the source term Jacobian. The properties of the Jacobian are very important for the stability of the method. We discuss what the chemical physics theory of chemical kinetics tells about the mathematical properties of the Jacobian matrix. We focus on the implication of the Le-Chatelier's principle on the sign of the diagonal entries of the Jacobian. We present the implementation of the method for turbulent flow. We use a two RANS turbulent model – one equation model – Spalart–Allmaras and a two-equation model – k–ω SST model. The last extension is for two-phase flows with a gas as a main phase and Eulerian representation of a dispersed particles phase (EDP). We present some examples for such flow computations inside a ballistic evaluation rocket motor. The numerical examples in this work include transonic flow about a RAE2822 airfoil, about a M6 Onera wing, NACA0012 airfoil at very low Mach number, two-phase flow inside a Ballistic evaluation motor (BEM), a turbulent reactive shear layer and a time dependent Sod's tube problem.

Original languageEnglish
Pages (from-to)201-234
Number of pages34
JournalJournal of Computational Physics
Volume358
DOIs
StatePublished - 1 Apr 2018

Keywords

  • Convergence acceleration
  • Multi-physics flows
  • Reactive flows
  • Turbulent flows

Fingerprint

Dive into the research topics of 'Acceleration methods for multi-physics compressible flow'. Together they form a unique fingerprint.

Cite this